Applied Machine Learning SoSe 2017
Contacts
Annalisa Marsico
Freie Universitaet Berlin, Max Planck Institute for Molecular Genetics
marsico@molgen.mpg.de, Annalisa.Marsico@fu-berlin.de
Bernhard Renard,
Robert Koch Insitute Berlin
RenardB@rki.de
Additional lecturer:
Thilo Muth,
Robert Koch Insitute Berlin
MuthT@rki.de
Robert Rentzsch
Robert Koch Institute Berlin
Rentzsch@rki.de
Roman Shulte-Sasse
Max Planck Institute for molecular Genetics Berlin
sasse@molgen.mpg.de
Prerequisites: Attended the Statistics course from the Master in Bioinformatics FU (or equivalent)
Maximum number of students: 15
Language: English
Time:
Lectures and Tutorials: Friday 10-14
Thanks for the enjoyable semester. Please note that there is no lecture on July 21st!
Location: 017/A6 (Arnimallee 6)
Goals
The students will be introduced to the basic statistical and algorithmic concepts in the field of Machine Learning, especially in the context of current research in bioinformatics, biology and biotechnology. This courses focuses on specific applications and data handling, rather than on theoretical concepts. Topics include:
- Regularization methods for variable selections and regression methods for features decorrelation with application to Mass Spectroscopy data and Cancer data
- Support Vector Machines for tumor classification based on genomic data and clinical covariates
- SVMs with string kernels to classify RNA sequences
- Artificial Neural Networks (ANNs) and Deep Learning and some recent applications in Bioinformatics
- Graphical models for signal cascade analysis and quasi-species identification
- Active learning with Random Forests applied to Mass Specrometry data
- Unsupervised learning: model-based clustering of microRNA expression data
Requirements
The students will be assigned weekly exercises which they have to complete. They will work on several practical problems and implement / use the methods learned during the lectures to extract information from biological datasets in R. Exercises are mandatory, problem sets will be posted on this website on a weekly basis and are to be handed in at the end of the Wednesday lecture.
Completing 80% of the assignment correctly and presenting in turns the results from the exercises, are prerequisites to pass the course.
More concretely, there are 13 tutorials and 80% of 13 is 10.4. Every assignment will be graded either ok, ok(-) and fail/NA
- "ok" referes to a good assignment with only minor errors
- "ok(-)" refers to non-failed assignment with major errors
- "fail" refers to a failed assignment; "NA" refers to a non-submitted assignment
Therefore the following scenarios are possible:
- If you get 13 ok(-) you pass
- If you get 12 ok(-) + 1 fail/NA you pass
- If you get 11 ok(-) + 2 fail/NA (and everything lower) you DON'T pass
- If you get 10 ok(-) +1 ok +2 fail/NA you pass
- Everything higher than this: you pass!
Grading criteria slides
Literature
Hastie, Tibshirani & Friedman. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer, 2009.
James, Witten, Hastie and Tibshirani. An introduciton to Statistical Learning (with applications in R). Springer, 2013
Christopher M. Bishop. Pattern Recognition and Machine Learning. Springer, 2006
Jon Shlens. A tutorial on Principal Component Analysis. 2003
Max Kuhn. Applied Predictive Modeling. Sprinder, 2013
Chapelle, Schoelkopf, Zien. Semi-supervised Learning, 2006
Resources on string kernels pdf1
Ian Goodfellow, Yoshua Bengio and Aaron Courville, Deep Learning
Lecture Materials
00 Introduction slides_Thilo, slides_Annalisa
01 Overview slides
02 Practical challenges in any ML study: data and benchmarking slides
03 Partial Least square Regression (PLSR) slides
04 Support Vector Machines (SVMs) slides
05 String kernel SVMs slides
06 Graphical Models slides
07 Nested Effects Models slides
08 Quasispecies Reconstruction slides
09 Artificial Neural Networks (ANNs) slides let's play with NNs
10 Deep Learning 1 slides
11 Deep Learning 2 slides
12 Active Learning slides
13 Regularisation slides
Tutorials